A Flexible Framework for Rogue Access Point Detection
Ricardo Gonçalves
1,2
, Manuel Eduardo Correia
1,2
and Pedro Brandão
1,3
1
CS Dept., Faculty of Sciences, University of Porto, Portugal
2
CRACS, INESC-TEC, Porto, Portugal
3
Instituto de Telecomunicações, Lisboa, Portugal
Keywords:
Wireless Networks, Rogue Access Points, Wireless Security, Detection Systems.
Abstract:
The society’s requirement for constant connectivity, leads to the need for an increasing number of available
Wi-Fi Access Points (
AP
s). These can be located almost everywhere: schools, coffee shops, shopping malls,
airports, trains, buses. This proliferation raises problems of trustworthiness and cost-effective difficulties for
verifying such security. In order to address these issues, it is necessary to detect effectively Rogue Access
Points (
RAP
s). There are open source solutions and others developed within enterprises for commercial
purposes. Relative to the latter, it has become obvious that they are not accessible to everyone due to their high
costs, and the former do not address all the types of
RAP
s. In this paper, we research these solutions and do a
thorough survey study of the most commonly used and recent Wi-Fi type of attacks. Based on this knowledge
we developed a solution to detect
RAP
s, which covers the most commonly known attacks. This proposed
solution, is a modular framework composed of Scanners, Detectors and Actuators, which are responsible for
scanning for available APs, apply a set of heuristics to detect them and apply a countermeasure mechanism.
1 INTRODUCTION
Today, The relevance of the Internet is something un-
matched with a past not so far away, where tasks were
mainly related to information search conducted via a
desktop computer. Currently, the tasks made using
the Internet evolved to banking, shopping, messaging,
video calls, gaming, social networks, geolocation, etc.
In short, a lot of essential day-to-day activities now re-
quire the Internet. In addition to the evolution in terms
of use and services, the Internet has also transformed
itself in the eyes of its users. So much so, that with the
arrival and massification of mobile devices, we have
witnessed a huge migration of network services usage
to wireless networks. This turned the Internet into
an almost ubiquitous service, with users demanding
access everywhere at any time. However, cost consi-
deration still plays a major role in how users choose to
connect to the Internet. Mobile data (cellular) contracts
are still expensive and capped, but available everyw-
here (indoor and outdoor). Free Wi-Fi access provided
by an Access Points (
AP
s), are generally indoors or in
otherwise selected public places. Conscientious users,
therefore, generally prefer Wi-Fi access (faster and
free) and only recur to mobile data when Wi-Fi access
is not immediately available or is of very poor qua-
lity. Mobile devices are even pre-configured to prefer
Wi-Fi access over mobile data.
In a generic scenario of mobile Internet usage se-
veral
AP
s are often used. But when a user requests a
wireless Internet connection, how can they be sure that
they are connecting to a trusted source/device?
In order to understand and answer this question,
our work addresses the problem of detecting false
Wi-Fi
AP
s that were not installed by an authorized
network administrator and can be used for malicious
purposes, thus becoming Rogue Access Points (
RAP
s).
The word "rogue" clarifies the malicious intentions of
this type of AP.
When an attacker sets up a
RAP
it will monitor
the traffic that goes through it and will be able to
perform several types of attacks, with Man in the
Middle (Schmoyer et al., 2004) being one of the most
popular.
In the first part of this paper we discuss some back-
ground knowledge needed to address this issue, spe-
cifically we focus on the types of
RAP
s and its more
popular types of attacks. Then, we present and dis-
cuss other related work in order to put into perspective
some approaches that have already been proposed in
the literature. In the second part of this paper we pre-
sent the framework architecture for our proposal and
explain in detail its design options and where it inno-
vates. We then finish by describing and discussing the
effectiveness of a proof of concept implementation of
our ideas that we deployed in the field.
466
Gonçalves, R., Correia, M. and Brandão, P.
A Flexible Framework for Rogue Access Point Detection.
DOI: 10.5220/0006832904660471
In Proceedings of the 15th International Joint Conference on e-Business and Telecommunications (ICETE 2018) - Volume 2: SECRYPT, pages 466-471
ISBN: 978-989-758-319-3
Copyright © 2018 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
2 BACKGROUND
In this section we start by presenting the types of
RAP
s
from the current literature. Then, we overview some
of the most common types of Wi-Fi attacks that can
be deployed from
RAP
s. Finally, we describe some
proposed countermeasures that can be used to prevent
or mitigate these attacks.
2.1 Types of RAPs
To classify the different types of
RAP
s, we divide them
into four general categories: Evil-Twin, Improperly
Configured, Unauthorized, and Compromised. This is
the nomenclature often used in the literature.
Evil Twin.
In the IEEE 802.11 standard (IEEE Wire-
less LAN Working Group, 2016) there are only two
identifiers for users to recognize an
AP
: the Service
Set Identifier (
SSID
) and the Basic Service Set Identi-
fier (
BSSID
). However, these identifiers can be easily
spoofed. The act of cloning a legitimate
AP
generates
an Evil Twin AP.
Evil Twin
RAP
s can exist in two forms: Coexis-
tence and Replacement. In the first, the legitimate AP
and the Evil Twin coexist in the same location. The
attacker increases the
RAP
s signal strength to force
users to connect to it, as the IEEE 802.11 standard sta-
tes that WLAN clients must connect to the
AP
that has
the strongest signal. In the Replacement type, the Evil
Twin replaces the legitimate
AP
by shutting it down,
using an active attack on it. To remain undetected to its
victims, the
RAP
needs to have an Internet connection
(or connectivity to the same network as the valid
AP
)
while in the first case it could relay the packets through
the legitimate AP, as long as it could connect to it.
Improperly Configured AP.
This type of
RAP
, as
its name suggests, is an
AP
which was improperly
configured. There is no adversary involved in the cre-
ation process. This can happen when, for instance,
an administrator does not use robust authentication
and encryption settings, leading to a network that can
be easily intruded.The
AP
s can also become vulne-
rable after software updates (Ma et al., 2008) or the
lack of recent updates. This type of
RAP
may lead to
backdoors in an organization’s network infrastructure.
Unauthorized AP.
This type of
RAP
is installed by
an employee or naïve user without the network admi-
nistrator’s permission (Whelan et al., 2011). This
RAP
is connected to the wired side of the network (like a
legitimate
AP
) and thus it is considered part of the
WLAN. A
RAP
of this kind is installed by uninformed
users for their own convenience, i.e., access to some
internal network resources, but it can also be deployed
with malicious intentions. Regardless, it will create
an unauthorized entry point into an organization’s net-
work and compromise its security.
Compromised AP.
When the shared keys used to
secure the network’s communications get compromi-
sed (Ma et al., 2008), the
AP
becomes a
RAP
. Thus,
the attacker that obtains the keys, will be able to join
the network.
2.2 Wi-Fi Attacks
RAP with Stronger Signal.
In this specific attack,
the
RAP
behaves as an Evil Twin. Here the attacker
entices the victim to connect to the Evil Twin AP.
For this attack to work, the attacker sets up the
wireless card into promiscuous or monitor mode in
order to gather information about the nearby
AP
s and
its clients. The
RAP
can now be created with the same
SSID
,
BSSID
and wireless medium physical channel,
and increasing the
AP
s signal power. Usually in this
type of attacks the Evil Twin
AP
is created with Open
authentication which eases its detection if the original
AP has some authentication enabled.
These two factors are paramount, because if no
authentication is used by the
RAP
, the signal strength
will not matter. A node will only roam when the
AP
s
has the same
SSID
, security mode and credentials.
Only if these conditions are met, can the target victim
be made to automatically switch to another
AP
with a
better signal strength.
De-authentication Attacks.
To enable roaming of
user, attackers often use de-authentication attacks to
force their victims to connect to the
RAP
. This attack
uses the special de-authentication frame of the 802.11
standard. The attacker sends it to the desired victim
on behalf of the legitimate
AP
and consequently the
victim will be de-authenticated from that
AP
. Then,
in the process of re-authentication, the victim will
automatically reconnect to the strongest AP.
Karma Attacks.
In these attacks, an automatic pro-
cess is used for cloning an AP, where the malicious
node listens for other wireless devices’ requests, and
creates a
RAP
s on demand. This is possible because of
a vulnerability in the method of discovering available
Wi-Fi networks (Dai Zovi and Macaulay, 2005).
When a client wants to join a Wi-Fi network, a scan
is made for the available networks (Probe request),
and then one network is selected, matching a Preferred
Network List (
PNL
). The attacker will listen for the
probe requests that are sent with the
SSID
s of the
client’s preferred networks. Thus, the attacker will be
able to respond to the probe request by sending a probe
response matching the network requested.
RAP with Radius Server.
This attack targets
WPA
Enterprise
AP
s. For this there are a couple of tools
A Flexible Framework for Rogue Access Point Detection
467
that can be used by an attacker:
hostapd-wpe
, and
freeradius-wpe
1
. These tools implement the IEEE
802.1X Authenticator and the Authentication Server
impersonation attacks in order to obtain client creden-
tials and establish connectivity to the client.
This
RAP
s taxonomy guarantees the scope of the
project and the possible types of attacks creates the
base cases for our proof of concept.
2.3 Countermeasures
This sub-section summarizes some countermeasures
and solutions to mitigate the described attacks.
We identify and differentiate between passive and
active countermeasures, and classify if the solution
can detect more than one type of
RAP
. There are
techniques that need protocol modifications and some
require the use of hardware. In passive techniques, the
detector radio listens on each channel for the periodic
beacons sent by an
AP
. While in actives, the detector
transmits a probe request (to inquire about available
WLANs/
SSID
s) and listens for a probe response from
an
AP
. However, a
RAP
usually does not reply to
active probing, so passive methods are often preferred.
The majority of techniques to detect
RAP
have focused on Evil Twin
RAP
types and
consequently will also detect
RAP
-based de-
authentication/disassociation attacks. There are
techniques to detect Unauthorized
AP
s (Whelan et al.,
2011), but practical techniques for the detection of
Compromised
AP
s are scarce. Moreover, there is no
single technique to detect all
RAP
types (Alotaibi and
Elleithy, 2016).
The techniques in Table 1 do not cover all the ty-
pes of
RAP
s and some of them even lack a proper
proof of concept tool for testing and benchmarking
purposes. However, they are the basis for the design
of some detection heuristics used in our work. During
our research, we have verified that there are several
freely available tools for deploying
RAP
s and perfor-
ming Wi-Fi attacks, e.g., Airbase-ng, hostapd-wpe,
PineAP, Pwnie Express, contrasting with the flagrant
lack of tools that can be employed to detect them, e.g.,
sentrygun
2
and EvilAP_Defender
3
.
1
https://github.com/OpenSecurityResearch/hostapd-
wpe and http://www.willhackforsushi.com/?page_id=37,
respectively.
2
https://github.com/s0lst1c3/sentrygun
3
https://github.com/moha99sa/EvilAP_Defender
Table 1: Detection techniques summary.
Protection
against Evil
Twin
Active Approach (Bratus et al.,
2008)
Timing-based (Han et al., 2011)
EAP-SWAT (Bauer et al., 2008)
(Han et al., 2011)
CETAD (Mustafa and Xu, 2014)
Unauthorized
countermeasu-
res
Unauthorized Approach (Yan
et al., 2009)
Agent-based (Chirumamilla and
Ramamurthy, 2003)
Protection
against
multiple RAPs
DWSA (Branch et al., 2004)
Hybrid RAP (Ma et al., 2008)
Multi-Agent (Sriram et al., 2010)
3 PROPOSED MODEL
The developed framework combines the most effective
techniques from the state of the art, and complements
them with new approaches to provide more accurate
results in the detection process. It is an open source
command line application
4
for Linux Systems, develo-
ped in Python, and with minimal dependencies.
This framework is a
RAP
Detector that works both
in a passive and active mode. It can be used by any
type of user, i.e., it can be configured with a profile of
a network administrator with knowledge about the in-
ternal network, or as a simple user with no knowledge
of the network (s)he is using.
The architecture is composed by the main applica-
tion, and it is connected to a set of modules: Scanners,
Detectors and Actuators.
Currently our detector node (a laptop running the
framework) is stationary. We are evaluating having
distributed nodes that send information to a central
server. The architecture of this central node would be
very similar to the one being presented here, where
some detectors and/or actuators modules would be
used to communicate with the remote nodes.
3.1 Profiles
The application is designed to use configuration files,
i.e., profiles. These profiles have information about
the network(s) the user connects to. There is a built in
profile for open/free
AP
s that are usually offered by
Internet Service Providers (
ISP
s)
5
. The user configu-
4
https://github.com/anotherik/RogueAP-Detector
5
Currently they are tuned for Portuguese ISPs.
SECRYPT 2018 - International Conference on Security and Cryptography
468
red profiles are optional and the application will run
without them. However, as expected this will generate
much less accurate results in the detection process,
since some heuristics will not be used.
3.2 Modules
The main application is connected to all the modules.
Scanners are passive modules with methods imple-
mented to monitor the network in order to find nearby
AP
s. After being started, the application will always be
scanning. The information provided will feed the De-
tectors, where a set of detection heuristics will sweep
the scanned
AP
s and when a specific
AP
reaches some
suspicion level it will interact with the Actuators to
perform active techniques.
For
Scanners
to scan the Wi-Fi channels we have
configured two different modules: one that uses the
iwlist
Linux command and another that uses the
scapy packet manipulation program.
On the
Detector
modules, one type of heuristic
present is the classic approach of whitelisting, where
we use the information from the configuration profile
regarding
SSID
s and their expected
BSSID
s. Since
this is likely to be bypassed (
BSSID
spoofing), we
also compare the Encryption type used by the AP.
Another heuristic is the variation of the signal
strength. As already described, this tool is designed
to be stationary, and normal
AP
s are also fixed. With
this in mind, we can estimate a baseline for the sig-
nal strength and use this information to improve the
detection process.
In particular the algorithm uses an
auth_rssi
de-
fined by the user for the authorized
AP
s
RSSI
, and
to avoid fluctuations a delta accounts for variations.
As such, the read RSSI value must fall in the allowed
range of
[auth_rssi delta;auth_rssi + delta]
. If this
is triggered, the user is prompted to associate to the
AP and continue to the active tests (Actuators).
The described heuristics assume that a configura-
tion profile is loaded when the application runs, ot-
herwise they will not be applied. In such case, a no
knowledge method is always run by the application,
where it conducts simple analyses on the scanned
AP
s.
A simple case is looking for
AP
s with the same
SSID
and different security being used.
Another heuristic used by the application is a free
Wi-Fi’s authenticity validation. This takes advantage
of the
BSSID
pattern of the studied free Wi-Fis. As
an example, the free Wi-Fi provided by the
ISP
NOS
(Portuguese media company), is generated from the
router of a personal network and the
BSSID
generated
for the free network follows a specific pattern. It is the
increment by one unit of the last byte of the personal
network’s
BSSID
. We use this information to analyse
the BSSIDs of the scanned free Wi-Fis.
This tool also has a blacklist passive heuristic for
RAP
s generated from known Wi-Fi attacking tools.
We configured two methods to detect Pineapple AP
RAP
s and
AP
s where the
BSSID
manufacturer is Alfa.
In the case of PineAP, the default
BSSID
contains
13:37
(leet speech), hence this heuristic can iden-
tify some
RAP
s configured by inexperienced attackers.
The Alfa condition is supported by the fact that these
Wi-Fi cards are mainly used for Wi-Fi attacks.
The passive detectors also have a heuristic to look
for de-authentication frames and alert the user. For
this we use the scapy packet manipulation tool.
The
Actuators
are composed by a set of active
detectors, a defensive mechanism and a honeypot.
The active detectors are performed when we have
knowledge about the scanned
AP
, i.e., the scanned
AP
matches an
AP
from our list of authorized
AP
s or for
open networks that fail the authenticity validation and
need further confirmation. The detectors under the
Actuators type of modules perform active operations
over the target
AP
, specifically, they associate to the
AP and then run the heuristics:
- Associate:
tries to associate to the AP and gather
information on the network addresses, comparing to
the values on the profile;
- Traceroute:
compares network paths to configured
destinations with the paths defined in the profile;
- Fingerprint:
verifies if the OS fingerprint and ser-
vices of the tested
AP
corresponds to the ones of an
authorized AP.
The Actuators set of modules is also composed
of an hybrid mechanism to detect Karma attacks. In
passive mode, we wait for probe responses from the
same
AP
to different probe requests. In the active
mode, we send probe requests for generated fictitious
SSID
s. If an
AP
sends responses for the fabricated
SSID
s, a Karma attack is spotted. In cases like this,
where it is obvious that
RAP
s are being created, a de-
fensive approach can be used by sending IEEE 802.11
de-authentication packets to the clients associated with
the detected RAP.
Last we have the honeypot module, which is a
RAP
created and controlled by the application. If an
attacker targets this
AP
, i.e., spoofing it, we have the
same confirmation as for the generated SSIDs.
4 RESULTS
This section describes the results obtained from the
RAP
detections performed on two different scenarios.
In the first, the
RAP
s were deployed by an information
A Flexible Framework for Rogue Access Point Detection
469
security specialist impersonating an attacker, where
we did not know the types of
RAP
. On the second,
the
RAP
s were configured and deployed by us on a
contained environment, to replicate all the attacks and
RAPs covered by the detection tool.
4.1 First Scenario - Unknown Setup
The tests from the first scenario were conducted in
an Enterprise environment. In such environments, the
Corporate
AP
s are usually configured with
WPA2
En-
terprise, which is associated with directory service
credentials to authenticate in the network.
Figure 1 has the first set of detections of the Rogue
AP
detector. In the screenshot the produced alert is
selected, and from the figure it is possible to under-
stand that an unauthorized
BSSID
was detected. The
network advertised by the RAP follows the alert.
Figure 1: Unknown scenario,
RAP
with same
SSID
and
different BSSID.
In this test case the
RAP
failed the
BSSID
parame-
ter which caused its discovery. To bypass this detection
the attacker modified the
BSSID
of the
RAP
to one
of the authorized list, but in order to succeed, i.e., to
have clients trying to associate with the
RAP
, the
RSSI
had to be increased. In this case, the tool detected the
signal strength strange behaviour and prompted the
user to associate to the
AP
. This association failed
corroborating that it was a RAP.
In this attack type, where a
RAP
is impersonating
a
WPA2
Enterprise network, the objective is to capture
the victim’s credentials, and for this it only requires
capturing the challenge/response authentication. This
means that usually, these attacks do not actually au-
thenticate the user with the RAP.
In the last test case of the first scenario we detect a
RAP
produced by a Pineapple
AP
with a Karma attack.
The attack used the default name of Pineapple and the
BSSID with 13:37, which our tool detects.
4.2 Second Scenario - Controlled Setup
For the second scenario, we only describe the results
for the heuristics not yet tested, namely:
RAP
with
different encryption, validation of free Wi-Fis authen-
ticity, and detection of de-authentication attacks.
Figure 2 shows the validated free Wi-Fis following
the algorithm described in section 3.2, and the case
of an Evil Twin attack where the
RAP
is configured
with different encryption. This type of detection is
also possible without using any profile, because of the
no knowledge detectors.
Another possible heuristic is the Timing Synchro-
nization Function (
TSF
) produced by
RAP
s. The
TSF
maintains the radios for all stations in the same Ba-
sic Service Set (
BSS
) synchronized. Each
AP
will
have a different value that is related with the up time
of the
AP
. In figure 2 it can be seen that the newly
created
RAP
has a
TSF
different from the authorized
AP
. From our analysis, we also verified that some
known
RAP
s tools have clearly different
TSF
values.
For example, the Airbase-ng
RAP
s are created with
extremely high values for this field, and
scapy
based
RAP
s use the zero value if not properly configured.
These values can be used for detection purposes.
In summary, we performed a set of tests for two
different scenarios, with the goal of creating a proof
of concept for the developed tool.
5 CONCLUSIONS
In this work we studied the concepts behind
RAP
s.
We described the types of
RAP
s and some counter-
measures studied in the literature. We also discussed
how these
RAP
s are created and explored. With this
knowledge, the analysis of Wi-Fi packets, and
AP
s
specifications and parameters, we proposed a flexible
framework that relies on modules and a set of heuris-
tics to detect possible
RAP
s. The modular architecture
eases the incorporation of new features, be it new scan-
ning methods or other detection heuristics.
The framework addresses: both Evil Twin types
of
RAP
s, coexistence and replacement, where it uses
the Actuators’ association procedure to identify them.
Improperly configured
AP
s, unauthorized and com-
promised
AP
s are also identified with whitelist and
blacklist heuristics. It is also possible to detect
RAP
s
generated by attacking tools, like PineAP, Alfa cards
and Airbase. A heuristic to validate authenticity of
free Wi-Fis is also configured in our application.
In the future we aim to expand the framework into
a distributed system where the main application would
communicate with a set of nodes. This would cover a
larger area, and ultimately, physically locate the
RAP
s.
For the scanning process, we would like to upgrade
iwlist
to
iw
, the latter has more information about
the Beacon frames that could be incorporated as new
detection heuristics. In the detection part we plan on
using the
TSF
parameter to discover newly created
SECRYPT 2018 - International Conference on Security and Cryptography
470
Figure 2: For controlled scenario Validate free Wi-Fis authenticity & –
RAP
with same
SSID
and
BSSID
, different Encryption.
AP
s and combine it with the other heuristics in order
to give a bigger accuracy in the detection rate.
ACKNOWLEDGEMENTS
This work is partially funded by the ERDF through the
COMPETE 2020 Programme within project POCI-01-
0145-FEDER-006961, and by National Funds through
the FCT as part of project UID/EEA/50014/2013.
REFERENCES
Alotaibi, B. and Elleithy, K. (2016). Rogue access point
detection: Taxonomy, challenges, and future directions.
Wireless Personal Communications, pages 1–30.
Bauer, K., Gonzales, H., and McCoy, D. (2008). Mitigating
evil twin attacks in 802.11. In 2008 IEEE Internati-
onal Performance, Computing and Communications
Conference, pages 513–516. IEEE.
Branch, J. W., Petroni, N. L., Doorn, L. V., and Safford,
D. (2004). Autonomic 802.11 wireless lan security
auditing. IEEE Security Privacy, 2(3):56–65.
Bratus, S., Cornelius, C., Kotz, D., and Peebles, D. (2008).
Active behavioral fingerprinting of wireless devices. In
Proceedings of the first ACM conference on Wireless
network security, pages 56–61. ACM.
Chirumamilla, M. K. and Ramamurthy, B. (2003). Agent
based intrusion detection and response system for wi-
reless lans. In IEEE International Conference on Com-
munications, ICC’03., volume 1, pages 492–496.
Dai Zovi, D. A. and Macaulay, S. A. (2005). Attacking
automatic wireless network selection. In Information
Assurance Workshop, 2005. IAW’05. Proceedings from
the Sixth Annual IEEE SMC, pages 365–372. IEEE.
Han, H., Sheng, B., Tan, C. C., Li, Q., and Lu, S.
(2011). A timing-based scheme for rogue ap detection.
IEEE Transactions on parallel and distributed Systems,
22(11):1912–1925.
IEEE Wireless LAN Working Group (2016). IEEE Standard
for Information technology–Telecommunications and
information exchange between systems Local and me-
tropolitan area networks–Specific requirements - Part
11: Wireless LAN Medium Access Control (MAC)
and Physical Layer (PHY) Specifications. IEEE Std
802.11-2016, pages 1–3534.
Ma, L., Teymorian, A. Y., and Cheng, X. (2008). A hybrid ro-
gue access point protection framework for commodity
wi-fi networks. In INFOCOM 2008. The 27th Confe-
rence on Computer Communications. IEEE. IEEE.
Mustafa, H. and Xu, W. (2014). Cetad: Detecting evil twin
access point attacks in wireless hotspots. In Commu-
nications and Network Security (CNS), 2014 IEEE
Conference on, pages 238–246. IEEE.
Schmoyer, T. R., Lim, Y. X., and Owen, H. L. (2004). Wire-
less intrusion detection and response: a classic study
using main-in-the-middle attack. In Wireless Commu-
nications and Networking Conference, 2004. WCNC.
2004 IEEE, volume 2, pages 883–888. IEEE.
Sriram, V. S., Sahoo, G., and Agrawal, K. K. (2010). De-
tecting and eliminating rogue access points in ieee-
802.11 wlan-a multi-agent sourcing methodology. In
Advance computing conference (IACC), 2010 IEEE
2nd international, pages 256–260. IEEE.
Whelan, R., Van Wagenen, L., and Morris, R. (2011). System
and method for detecting unauthorized wireless access
points. US Patent 7,965,842.
Yan, B., Chen, G., Wang, J., and Yin, H. (2009). Robust de-
tection of unauthorized wireless access points. Mobile
Networks and Applications, 14(4):508–522.
A Flexible Framework for Rogue Access Point Detection
471